On Saturday 18 April 2009 12:25:17 ext Jari Arkko wrote: > When the work started, it was clearly about multiple interfaces. Upon > closer inspection, we have realized that the overall problem is somewhat > larger. Problems that occur with multiple interfaces also occur even > with one interface, when you have a number of default routers on the > same link. The current charter text reflects this in some parts of the > text, e.g., I don't quite agree here. Multiple interfaces will typically have some differing properties, such as: - bandwidth, - latency, - charging, - power consumption, - reliability, - address transparency (e.g. what kind of NAT if any) - filtering policy, - IP protocol versions availability, - set of reachable hosts (the Internet, the home, the office...), - recursive DNS servers, - DNS namespace, - address filtering rules. A host may have multiple default gateways on the same "interface". However, those gateways should be usable interchangeably. All of those gateways should have the same "values" for the properties above. Otherwise, I would argue the network is broken and/or the host is misconfigured. In other words, it can pick any of its live gateways outbound, and may receive packets from any of those gateways in the other direction. If I'm not mistaken, the IETF IPv6 wireless network is an example of this. But I think /that/ form of multi-homing has been solved for a long time by Ip stack implementors. What was discussed at the MIF BoF were quite different scenarios. -- Rémi Denis-Courmont Nokia Devices R&D, Maemo Software, Helsinki _______________________________________________ Ietf@xxxxxxxx https://www.ietf.org/mailman/listinfo/ietf